[
https://issues.apache.org/jira/browse/STORM-193?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13873260#comment-13873260
]
Gareth Daniel Smith commented on STORM-193:
-------------------------------------------
Note that there appears to be a very similar problem with the
backtype.storm.LocalCluster.shutdown() method - it is not synchronous either.
I have worked around this issue by sleeping for a few seconds in my integration
tests, but this is not robust and makes the tests take longer than necessary.
I think it would be good if LocalCluster had a method to call that doesn't
return until the cluster is completely shutdown.
> I think it would be more user friendly if [storm kill ...] was synchronous
> --------------------------------------------------------------------------
>
> Key: STORM-193
> URL: https://issues.apache.org/jira/browse/STORM-193
> Project: Apache Storm (Incubating)
> Issue Type: Wish
> Reporter: Gareth Daniel Smith
> Priority: Minor
>
> Currently the [storm kill topology-name] command is not synchronous, even if
> [-w 0] is used. After killing a topology using the kill command then one must
> wait an unspecified amount of time (a second or so) before running [storm
> jar] with the same topology name, otherwise storm will complain that the
> topology already exists.
> This is frustrating me during development when I am repeatedly killing and
> deploying new versions of my topology.
> I would like if there was an option to have [storm kill] be synchronous, or
> even if it was the default behavior.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
